A bug in some Ray-Ban Meta smart glasses allowed the Meta AI assistant to activate, capture audio and generate transcripts without the wearer saying “Hey Meta”—even when the setting that permits wake-word-free activation was switched off. Meta says the problem has been resolved. What the company has not supplied is an explanation of what went wrong, how many devices were affected, which firmware version contains the fix, or what happens to the audio that was recorded by mistake.
- What happened: One owner saw the behavior on three pairs of second- and third-generation glasses after a firmware update.
- Meta’s response: It says the bug is fixed and users can delete voice logs manually, but it has not explained the cause, scope or firmware fix.
- Why it matters: Meta already stores voice recordings by default and says some audio may be reviewed by staff.
What the glasses actually did
The reports describe an accidental assistant activation, not a continuous covert listening mode. False positives are a known problem for voice assistants; the difference here is the hardware. These glasses are worn during dinners, meetings and private conversations, with a microphone on the wearer’s face.
According to the user, conversations were recorded and transcribed into the voice activity log in the Meta AI app. In some cases, the white activation LED reportedly turned on during ordinary conversation. The outlet that first reported the case could not independently reproduce it and described the reports as limited.
Meta’s short answer leaves the hard questions open
Meta’s response was brief. A spokesperson said:
“This issue has been resolved. People can manage, review and delete voice logs of their interactions with Meta AI via settings at any time.”
That confirms a manual deletion path, not an explanation. Meta has not said what caused the bug, whether a faulty update was responsible, how many users were affected, or when the fix was deployed. It has also not explained what happens to audio already captured. The only remedy offered is for the user to find and delete it.
Support guidance reported after the case was even blunter: users were advised to disable microphone and camera permissions, turn off usage detection, clear their history and switch off automatic updates. Together, those steps neutralize much of the product’s functionality—a workaround that shifts the burden onto the customer.
Default logging, legal risk and a wider wearable problem
The episode is complicated by Meta’s existing practices. The company already stored voice recordings from these glasses by default, and its privacy notice mentions unintentional activations and the possibility that some material may be reviewed by personnel to improve voice services. Earlier, Meta removed the option to avoid storing recordings at all. A bug that triggers the assistant has very different consequences depending on what the system does with the resulting audio—and Meta has not clarified that chain here.
Regulators are drawing boundaries. In Spain, the AEPD reminded users this summer that voice recordings of identified or identifiable people are personal data and that responsibility for misuse falls on the wearer. The agency also said smart glasses can have legitimate uses and do not automatically violate data-protection rules. No authority has ruled on this incident, and no affected users in the European Union have been identified.
The user who raised the alarm is consulting a lawyer about a possible class action if other customers find similar recordings. No lawsuit has been filed. Meta is already facing separate scrutiny: a class action in Chicago accuses it of developing facial recognition for the glasses, and a case filed in March alleges footage recorded on the devices was shared with contractors in Kenya. Several UK pubs have also banned the glasses because people did not reliably notice the recording indicator.
The bug lands as the wearable industry moves toward always-listening AI devices. Amazon recently launched the Bee Frontier, a $74.99 bracelet, and the Bee Eclipse, a $99.99 magnetic wearable for MagSafe phones. Both continuously listen, transcribe and summarize conversations. Legal experts warn such products could violate state wiretap laws requiring consent from every party in a conversation—a standard bystanders have not met.
Meta’s technical fix may be real. The reputational problem is not. For a device sold as a social accessory but carrying a microphone into other people’s space, transparency is part of the safety architecture. A two-line response, with no cause, scope or data-handling detail, leaves users to audit their own logs and regulators to fill in the blanks. That is a fragile foundation for a product category that depends on trust.